O N L I N E a N O M a L Y D E T E C T I O N

Where the ro le o f so f twarein tens ive sys tems has sh i f ted f rom the t rad i t iona l one o f fu l f i l l ing i so lated computat iona l t a sks , l a rger co l l aborat ive soc ie t ie s wi th interac t ion as pr imary resource, i s g radua l ly t ak ing i t s p lace. Thi s can be obser ved in anyth ing f rom log i s t i cs to rescue operat ions and resource management , numerous ser v ices wi th key-ro les in the moder n in f ra s t r ucture. In the l ight o f th i s new co l laborat ive order, i t i s imperat ive that the too l s (compi ler s , debugger s , prof i l er s ) and methods ( requ irements , des ign, implementat ion, te s t ing) that suppor t t rad i t iona l so f tware eng ineer ing va lues a l so ad jus t and extend towards those nur tured by the onl ine in s t r umentat ion o f sof tware intens ive sys tems. That i s , to ad jus t and he lp to avo id s i tuat ions where l imi tat ions in tec hnology and methodology would prevent us f rom ascer ta in ing the wel l -be ing and secur i ty o f sys tems that a s s i s t s our very l ives. Coupled wi th most per spect ives on so f tware development and ma intenance i s one wel l e s t abl i shed member o f , and complement to, the deve lopment process. Debugg ing; or the ar t o f d i scover ing, loca l i s ing, and cor rect ing undes i rable behaviour s in so f twarein tens ive sys tems, the need for whic h tend to f ar out l i ve deve l opment in i t se l f . Debugg ing i s cur rent ly per for med based on a premise o f the developer operat ing f rom a godl ike per spect ive. A per spect ive that impl ie s acces s and knowledge regarding source code, a long wi th minute contro l over execut ion proper t ie s. However, the qua l i ty a s wel l a s acces s ib i l i ty o f suc h in for mat ion s tead i ly dec l ine wi th t ime a s requ irements , implementat ion, hardware components and the i r as soc i ated deve loper s , a l l a l ike fa l l beh ind the i r cont inuous ly evolv ing sur roundings. In th i s thes i s i t i s argued that the cur rent pract ice of so f tware debugg ing i s insu f f i c ient , and as precur sory act ion, in troduce a tec hn ica l p lat for m su i t able for exper i ment ing with future methods regard ing on l ine debugg ing, ma intenance and ana lys i s. An in i t i a l implementat ion o f th i s p lat for m wi l l then be used for exper iment ing wi th a s imple method that i s ta rget ing on l ine obser vat ion of sof tware behav iour.